GRIDEX is the first pipeline to generate grid-grounded structured forensic explanations for deepfake spectrogram anomalies using a two-stage SFT plus GRPO training process on vision-language models.
Interpretable all-type audio deepfake detection with audio llms via frequency-time reinforcement learning
6 Pith papers cite this work. Polarity classification is still indexing.
6
Pith papers citing it
citation-role summary
background 2
method 1
citation-polarity summary
fields
cs.SD 6years
2026 6verdicts
UNVERDICTED 6representative citing papers
A gated fusion of XLSR-53 and CORES features with energy margin and diversity losses reaches 97.6% ID accuracy and reduces FPR95 by 83.5% relative to the Interspeech 2025 baseline on MLAAD.
APEX generates four types of prototype-based explanations for pre-trained audio classifiers that preserve output invariance and target acoustic properties better than gradient methods applied to spectrograms.
A survey of Large Audio Language Models that establishes a taxonomy of trustworthiness vulnerabilities and proposes a Defense-in-Depth roadmap for audio intelligence.
EnvTriCascade is a tri-stage cascaded framework using mix-consistency detection followed by dual SSL-based five-class classifiers with cross-branch attention and RawBoost augmentation, achieving 0.8266 Macro-F1 on the ESDD2 2026 challenge test set.
AT-ADD introduces standardized tracks and datasets for evaluating audio deepfake detectors on speech under real-world conditions and on diverse unknown audio types to promote generalization beyond speech-centric methods.
citing papers explorer
-
GRIDEX: Grid-Grounded Forensic Explanations for Deepfake Spectrogram Analysis
GRIDEX is the first pipeline to generate grid-grounded structured forensic explanations for deepfake spectrogram anomalies using a two-stage SFT plus GRPO training process on vision-language models.
-
Dual-Branch Gated Fusion for Open-Set Audio Deepfake Source Tracing
A gated fusion of XLSR-53 and CORES features with energy margin and diversity losses reaches 97.6% ID accuracy and reduces FPR95 by 83.5% relative to the Interspeech 2025 baseline on MLAAD.
-
APEX: Audio Prototype EXplanations for Classification Tasks
APEX generates four types of prototype-based explanations for pre-trained audio classifiers that preserve output invariance and target acoustic properties better than gradient methods applied to spectrograms.
-
A Survey of Large Audio Language Models: Generalization, Trustworthiness, and Outlook
A survey of Large Audio Language Models that establishes a taxonomy of trustworthiness vulnerabilities and proposes a Defense-in-Depth roadmap for audio intelligence.
-
EnvTriCascade: An Environment-Aware Tri-Stage Cascaded Framework for ESDD2 2026 Challenge
EnvTriCascade is a tri-stage cascaded framework using mix-consistency detection followed by dual SSL-based five-class classifiers with cross-branch attention and RawBoost augmentation, achieving 0.8266 Macro-F1 on the ESDD2 2026 challenge test set.
-
AT-ADD: All-Type Audio Deepfake Detection Challenge Evaluation Plan
AT-ADD introduces standardized tracks and datasets for evaluating audio deepfake detectors on speech under real-world conditions and on diverse unknown audio types to promote generalization beyond speech-centric methods.